diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index 2ce59b6aad..77cd1d5200 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -6,6 +6,9 @@ on:
     tags: [v0.*]
   pull_request:
 
+env:
+  RUST_BACKTRACE: 1
+
 jobs:
   build:
     strategy:
@@ -18,6 +21,13 @@ jobs:
             channel: stable
             target: x86_64-pc-windows-msvc
             kind: test
+            backends: dx12 # dx11
+
+          - name: Windows Nightly x86_64
+            os: windows-2019
+            channel: nightly
+            target: x86_64-pc-windows-msvc
+            kind: compile
           
           # MacOS
 
@@ -42,7 +52,8 @@ jobs:
             os: ubuntu-20.04
             channel: stable
             target: x86_64-unknown-linux-gnu
-            kind: test
+            kind: compile
+            backends: # vulkan gl
 
           - name: Linux Nightly x86_64
             os: ubuntu-20.04
@@ -71,7 +82,7 @@ jobs:
 
     env:
       PKG_CONFIG_ALLOW_CROSS: 1 # allow android to work
-      RUSTFLAGS: --cfg=web_sys_unstable_apis
+      RUSTFLAGS: --cfg=web_sys_unstable_apis -D warnings
       RUSTDOCFLAGS: -Dwarnings
 
     steps:
@@ -87,6 +98,11 @@ jobs:
           profile: minimal
           components: clippy
 
+      - name: caching
+        uses: Swatinem/rust-cache@v1
+        with:
+          key: ${{ matrix.target }}-a # suffix for cache busting
+
       - name: add android apk to path
         if: matrix.os == 'ubuntu-20.04' && matrix.target == 'aarch64-linux-android'
         run: |
@@ -95,73 +111,69 @@ jobs:
       - name: install llvmpipe and lavapipe
         if: matrix.os == 'ubuntu-20.04' && matrix.kind == 'test'
         run: |
-          echo "Installing Vulkan"
           sudo apt-get update -y -qq
           sudo add-apt-repository ppa:ubuntu-x-swat/updates -y
           sudo apt-get update
           sudo apt install -y libxcb-xfixes0-dev mesa-vulkan-drivers
+        # add libegl1-mesa libgl1-mesa-dri for egl
+
+      # We enable line numbers for panics, but that's it
+      - name: disable debug
+        shell: bash
+        run: |
+          mkdir .cargo
+          echo """[profile.dev]
+          debug = 1" > .cargo/config.toml
 
       # This is separate for now because webgl isn't hooked up so we can't compile wgpu-core for wasm
       - name: check web
         if: matrix.kind == 'webgl'
         run: |
-          cargo clippy --target ${{ matrix.target }} -p wgpu -- -D warnings
+          cargo clippy --target ${{ matrix.target }} -p wgpu
 
           # build docs
           cargo doc --target ${{ matrix.target }} -p wgpu --no-deps
 
-      - name: check native stable (fatal warnings)
-        if: (matrix.kind == 'compile' || matrix.kind == 'test') && matrix.channel == 'stable'
-        run: |
-          # check with no features
-          cargo clippy --target ${{ matrix.target }} -p wgpu -p wgpu-core -p wgpu-info -p player -- -D warnings
-
-          # check with all features
-          # explicitly don't mention wgpu-hal so that --all-features don't apply to it
-          cargo clippy --target ${{ matrix.target }} -p wgpu -p wgpu-core -p wgpu-info -p player --examples --tests --all-features -- -D warnings 
-
-          # build docs
-          cargo doc --target ${{ matrix.target }} --no-deps
-          cargo doc --target ${{ matrix.target }} -p wgpu -p wgpu-core -p wgpu-info -p player --all-features --no-deps
-
-      - name: check native nightly (non-fatal warnings)
-        if: (matrix.kind == 'compile' || matrix.kind == 'test') && matrix.channel != 'stable'
+      - name: check native
+        if: matrix.kind == 'compile' || matrix.kind == 'test'
         run: |
           # check with no features
           cargo clippy --target ${{ matrix.target }} -p wgpu -p wgpu-core -p wgpu-info -p player
 
           # check with all features
           # explicitly don't mention wgpu-hal so that --all-features don't apply to it
-          cargo clippy --target ${{ matrix.target }} -p wgpu -p wgpu-core -p wgpu-info -p player --examples --tests --all-features
+          cargo clippy --target ${{ matrix.target }} -p wgpu -p wgpu-core -p wgpu-info -p player --examples --tests --all-features 
 
-      - name: build tests
-        if: matrix.kind == 'test'
-        run: |
-          cargo build --target ${{ matrix.target }} --bin wgpu-info 
-          cargo build --target ${{ matrix.target }} --tests -p wgpu-types -p wgpu-hal -p wgpu-core
-          cargo build --target ${{ matrix.target }} --examples --tests -p wgpu
+          # build docs
+          cargo doc --target ${{ matrix.target }} --no-deps
+          cargo doc --target ${{ matrix.target }} -p wgpu -p wgpu-core -p wgpu-info -p player --all-features --no-deps
 
       - name: tests
-        if: matrix.kind == 'test' && matrix.os == 'windows-2019'
+        if: matrix.kind == 'test'
+        shell: bash
         run: |
           # run wgpu-info
           cargo run --target ${{ matrix.target }} --bin wgpu-info
-          # run player tests
-          cargo test --target ${{ matrix.target }} -p wgpu-types -p wgpu-hal -p wgpu-core --no-fail-fast -- --nocapture
-          # run coretests
-          cargo run --target ${{ matrix.target }} --bin wgpu-info -- cargo test --target ${{ matrix.target }} -p wgpu --no-fail-fast -- --nocapture
+          for backend in ${{ matrix.backends }}; do
+            echo "======= NATIVE TESTS $backend ======";
+            # run player tests
+            WGPU_BACKEND=$backend cargo test --target ${{ matrix.target }} -p wgpu-types -p wgpu-hal -p wgpu-core --no-fail-fast -- --nocapture --test-threads=1
+            # run coretests
+            WGPU_BACKEND=$backend cargo test --target ${{ matrix.target }} -p wgpu --no-fail-fast -- --nocapture --test-threads=1
+          done
 
-  test:
+  cts:
     strategy:
       fail-fast: false
       matrix:
         include:
           # Windows
           - name: Windows x86_64
-            os: windows-latest
+            os: windows-2019
             target: x86_64-pc-windows-msvc
+            backends: dx12 # dx11
 
-    name: Test ${{ matrix.name }}
+    name: CTS ${{ matrix.name }}
     runs-on: ${{ matrix.os }}
 
     steps:
@@ -169,6 +181,7 @@ jobs:
         uses: actions/checkout@v2
         with:
           path: wgpu
+      
       - name: checkout cts
         uses: actions/checkout@v2
         with:
@@ -183,16 +196,36 @@ jobs:
           target: ${{ matrix.target }}
           profile: minimal
 
+      - name: caching
+        uses: Swatinem/rust-cache@v1
+        with:
+          key: cts-a # suffix for cache busting
+          working-directory: wgpu/cts_runner
+          target-dir: wgpu/cts_runner/target
+
+      # We enable line numbers for panics, but that's it
+      - name: disable debug
+        shell: bash
+        run: |
+          mkdir wgpu/.cargo
+          echo """[profile.dev]
+          debug = 1" > wgpu/.cargo/config.toml
+
       - name: build CTS runner
-        run: cd wgpu/cts_runner && cargo build
+        run: |
+          cargo build --manifest-path wgpu/cts_runner/Cargo.toml 
 
       - name: run CTS
+        if: ${{ matrix.os }} == 'windows-2019'
         shell: bash
-        run: >-
+        run: |
           cd cts;
-          grep -v '^//' ../wgpu/cts_runner/test.lst | while IFS=$' \t\r\n' read test; do
-            echo "=== Running $test ===";
-            cargo run --manifest-path ../wgpu/cts_runner/Cargo.toml --frozen -- ./tools/run_deno --verbose "$test";
+          for backend in ${{ matrix.backends }}; do
+            echo "======= CTS TESTS $backend ======";
+            grep -v '^//' ../wgpu/cts_runner/test.lst | while IFS=$' \t\r\n' read test; do
+              echo "=== Running $test ===";
+              DENO_WEBGPU_BACKEND=$backend cargo run --manifest-path ../wgpu/cts_runner/Cargo.toml --frozen -- ./tools/run_deno --verbose "$test";
+            done
           done
 
   fmt:
diff --git a/deno_webgpu/src/lib.rs b/deno_webgpu/src/lib.rs
index 3e1f18b572..d8c8b67e98 100644
--- a/deno_webgpu/src/lib.rs
+++ b/deno_webgpu/src/lib.rs
@@ -242,13 +242,14 @@ pub async fn op_webgpu_request_adapter(
 ) -> Result<GpuAdapterDeviceOrErr, AnyError> {
     let mut state = state.borrow_mut();
     check_unstable(&state, "navigator.gpu.requestAdapter");
+    let backends = std::env::var("DENO_WEBGPU_BACKEND").ok().map_or_else(wgpu_types::Backends::all, |s| wgpu_core::instance::parse_backends_from_comma_list(&s));
     let instance = if let Some(instance) = state.try_borrow::<Instance>() {
         instance
     } else {
         state.put(wgpu_core::hub::Global::new(
             "webgpu",
             wgpu_core::hub::IdentityManagerFactory,
-            wgpu_types::Backends::PRIMARY,
+            backends,
         ));
         state.borrow::<Instance>()
     };
@@ -263,7 +264,7 @@ pub async fn op_webgpu_request_adapter(
     };
     let res = instance.request_adapter(
         &descriptor,
-        wgpu_core::instance::AdapterInputs::Mask(wgpu_types::Backends::PRIMARY, |_| {
+        wgpu_core::instance::AdapterInputs::Mask(backends, |_| {
             std::marker::PhantomData
         }),
     );
diff --git a/wgpu-core/src/binding_model.rs b/wgpu-core/src/binding_model.rs
index e745374ffe..e2c7472c63 100644
--- a/wgpu-core/src/binding_model.rs
+++ b/wgpu-core/src/binding_model.rs
@@ -419,6 +419,7 @@ pub struct BindGroupLayout<A: hal::Api> {
     pub(crate) device_id: Stored<DeviceId>,
     pub(crate) multi_ref_count: MultiRefCount,
     pub(crate) entries: BindEntryMap,
+    #[allow(unused)]
     pub(crate) dynamic_count: usize,
     pub(crate) count_validator: BindingTypeMaxCountValidator,
     #[cfg(debug_assertions)]
diff --git a/wgpu-core/src/device/mod.rs b/wgpu-core/src/device/mod.rs
index bf04019aaf..277a28cc6c 100644
--- a/wgpu-core/src/device/mod.rs
+++ b/wgpu-core/src/device/mod.rs
@@ -4637,7 +4637,7 @@ impl<G: GlobalIdentityHandlerFactory> Global<G> {
                     resource::BufferMapState::Waiting(resource::BufferPendingMapping {
                         range,
                         op,
-                        parent_ref_count: buffer.life_guard.add_ref(),
+                        _parent_ref_count: buffer.life_guard.add_ref(),
                     })
                 }
             };
diff --git a/wgpu-core/src/hub.rs b/wgpu-core/src/hub.rs
index 4240ecdee3..d43c68a7de 100644
--- a/wgpu-core/src/hub.rs
+++ b/wgpu-core/src/hub.rs
@@ -17,21 +17,12 @@ use std::cell::Cell;
 use std::{fmt::Debug, marker::PhantomData, mem, ops};
 
 /// A simple structure to manage identities of objects.
-#[derive(Debug)]
+#[derive(Debug, Default)]
 pub struct IdentityManager {
     free: Vec<Index>,
     epochs: Vec<Epoch>,
 }
 
-impl Default for IdentityManager {
-    fn default() -> Self {
-        Self {
-            free: Default::default(),
-            epochs: Default::default(),
-        }
-    }
-}
-
 impl IdentityManager {
     pub fn from_index(min_index: u32) -> Self {
         Self {
diff --git a/wgpu-core/src/instance.rs b/wgpu-core/src/instance.rs
index 6407259a5a..7389a79b8e 100644
--- a/wgpu-core/src/instance.rs
+++ b/wgpu-core/src/instance.rs
@@ -876,3 +876,40 @@ impl<G: GlobalIdentityHandlerFactory> Global<G> {
         (id, Some(error))
     }
 }
+
+/// Generates a set of backends from a comma separated list of case-insensitive backend names.
+///
+/// Whitespace is stripped, so both 'gl, dx12' and 'gl,dx12' are valid.
+///
+/// Always returns WEBGPU on wasm over webgpu.
+///
+/// Names:
+/// - vulkan = "vulkan" or "vk"
+/// - dx12   = "dx12" or "d3d12"
+/// - dx11   = "dx11" or "d3d11"
+/// - metal  = "metal" or "mtl"
+/// - gles   = "opengl" or "gles" or "gl"
+/// - webgpu = "webgpu"
+pub fn parse_backends_from_comma_list(string: &str) -> Backends {
+    let mut backends = Backends::empty();
+    for backend in string.to_lowercase().split(',') {
+        backends |= match backend.trim() {
+            "vulkan" | "vk" => Backends::VULKAN,
+            "dx12" | "d3d12" => Backends::DX12,
+            "dx11" | "d3d11" => Backends::DX11,
+            "metal" | "mtl" => Backends::METAL,
+            "opengl" | "gles" | "gl" => Backends::GL,
+            "webgpu" => Backends::BROWSER_WEBGPU,
+            b => {
+                log::warn!("unknown backend string '{}'", b);
+                continue;
+            }
+        }
+    }
+
+    if backends.is_empty() {
+        log::warn!("no valid backend strings found!");
+    }
+
+    backends
+}
diff --git a/wgpu-core/src/present.rs b/wgpu-core/src/present.rs
index ab25b8afc2..9daecf326b 100644
--- a/wgpu-core/src/present.rs
+++ b/wgpu-core/src/present.rs
@@ -32,6 +32,7 @@ pub const DESIRED_NUM_FRAMES: u32 = 3;
 pub(crate) struct Presentation {
     pub(crate) device_id: Stored<DeviceId>,
     pub(crate) config: wgt::SurfaceConfiguration,
+    #[allow(unused)]
     pub(crate) num_frames: u32,
     pub(crate) acquired_texture: Option<Stored<TextureId>>,
 }
diff --git a/wgpu-core/src/resource.rs b/wgpu-core/src/resource.rs
index 4e9d971e4c..9f990b995a 100644
--- a/wgpu-core/src/resource.rs
+++ b/wgpu-core/src/resource.rs
@@ -109,7 +109,7 @@ pub(crate) struct BufferPendingMapping {
     pub range: Range<wgt::BufferAddress>,
     pub op: BufferMapOperation,
     // hold the parent alive while the mapping is active
-    pub parent_ref_count: RefCount,
+    pub _parent_ref_count: RefCount,
 }
 
 pub type BufferDescriptor<'a> = wgt::BufferDescriptor<Label<'a>>;
diff --git a/wgpu-core/src/validation.rs b/wgpu-core/src/validation.rs
index 43d000c0e4..e1e024b117 100644
--- a/wgpu-core/src/validation.rs
+++ b/wgpu-core/src/validation.rs
@@ -21,6 +21,7 @@ enum ResourceType {
 
 #[derive(Debug)]
 struct Resource {
+    #[allow(unused)]
     name: Option<String>,
     bind: naga::ResourceBinding,
     ty: ResourceType,
@@ -90,6 +91,7 @@ enum Varying {
     BuiltIn(naga::BuiltIn),
 }
 
+#[allow(unused)]
 #[derive(Debug)]
 struct SpecializationConstant {
     id: u32,
@@ -101,6 +103,7 @@ struct EntryPoint {
     inputs: Vec<Varying>,
     outputs: Vec<Varying>,
     resources: Vec<(naga::Handle<Resource>, GlobalUse)>,
+    #[allow(unused)]
     spec_constants: Vec<SpecializationConstant>,
     sampling_pairs: FastHashSet<(naga::Handle<Resource>, naga::Handle<Resource>)>,
 }
diff --git a/wgpu-hal/src/dx12/mod.rs b/wgpu-hal/src/dx12/mod.rs
index 6e4a0b4cbe..886e3807f6 100644
--- a/wgpu-hal/src/dx12/mod.rs
+++ b/wgpu-hal/src/dx12/mod.rs
@@ -155,13 +155,17 @@ unsafe impl Sync for Surface {}
 
 #[derive(Debug, Clone, Copy)]
 enum MemoryArchitecture {
-    Unified { cache_coherent: bool },
+    Unified {
+        #[allow(unused)]
+        cache_coherent: bool,
+    },
     NonUnified,
 }
 
 #[derive(Debug, Clone, Copy)]
 struct PrivateCapabilities {
     instance_flags: crate::InstanceFlags,
+    #[allow(unused)]
     heterogeneous_resource_heaps: bool,
     memory_architecture: MemoryArchitecture,
     heap_create_not_zeroed: bool,
diff --git a/wgpu-hal/src/gles/egl.rs b/wgpu-hal/src/gles/egl.rs
index 7badd3c6d0..c91c1a2d3d 100644
--- a/wgpu-hal/src/gles/egl.rs
+++ b/wgpu-hal/src/gles/egl.rs
@@ -317,6 +317,7 @@ impl AdapterContext {
 #[derive(Debug)]
 struct Inner {
     egl: Arc<egl::DynamicInstance<egl::EGL1_4>>,
+    #[allow(unused)]
     version: (i32, i32),
     supports_native_window: bool,
     display: egl::Display,
@@ -721,6 +722,7 @@ pub struct Swapchain {
     extent: wgt::Extent3d,
     format: wgt::TextureFormat,
     format_desc: super::TextureFormatDesc,
+    #[allow(unused)]
     sample_type: wgt::TextureSampleType,
 }
 
@@ -731,6 +733,7 @@ pub struct Surface {
     config: egl::Config,
     display: egl::Display,
     context: egl::Context,
+    #[allow(unused)]
     pbuffer: Option<egl::Surface>,
     pub(super) presentable: bool,
     raw_window_handle: RawWindowHandle,
diff --git a/wgpu-hal/src/gles/mod.rs b/wgpu-hal/src/gles/mod.rs
index 5c4e752682..631d86eb4d 100644
--- a/wgpu-hal/src/gles/mod.rs
+++ b/wgpu-hal/src/gles/mod.rs
@@ -229,6 +229,7 @@ pub struct Texture {
     mip_level_count: u32,
     array_layer_count: u32,
     format: wgt::TextureFormat,
+    #[allow(unused)]
     format_desc: TextureFormatDesc,
     copy_size: crate::CopyExtent,
 }
@@ -345,6 +346,7 @@ struct VertexBufferDesc {
     stride: u32,
 }
 
+#[allow(unused)]
 #[derive(Clone)]
 struct UniformDesc {
     location: glow::UniformLocation,
@@ -533,6 +535,7 @@ enum Command {
         dst_target: BindTarget,
         range: crate::MemoryRange,
     },
+    #[allow(unused)] // yet unimplemented
     ClearTexture {
         dst: glow::Texture,
         dst_target: BindTarget,
@@ -554,6 +557,7 @@ enum Command {
     },
     CopyBufferToTexture {
         src: glow::Buffer,
+        #[allow(unused)]
         src_target: BindTarget,
         dst: glow::Texture,
         dst_target: BindTarget,
@@ -565,6 +569,7 @@ enum Command {
         src_target: BindTarget,
         src_format: wgt::TextureFormat,
         dst: glow::Buffer,
+        #[allow(unused)]
         dst_target: BindTarget,
         copy: crate::BufferTextureCopy,
     },
diff --git a/wgpu-types/src/lib.rs b/wgpu-types/src/lib.rs
index e59f9e67a2..efc8125e8d 100644
--- a/wgpu-types/src/lib.rs
+++ b/wgpu-types/src/lib.rs
@@ -3,11 +3,6 @@
  */
 
 #![allow(
-    // The intra doc links to the wgpu crate in this crate actually successfully link to the types in the wgpu crate, when built from the wgpu crate.
-    // However when building from both the wgpu crate or this crate cargo doc will claim all the links cannot be resolved
-    // despite the fact that it works fine when it needs to.
-    // So we just disable those warnings.
-    broken_intra_doc_links,
     // We don't use syntax sugar where it's not necessary.
     clippy::match_like_matches_macro,
 )]
@@ -722,6 +717,8 @@ impl Limits {
 #[derive(Clone, Debug, PartialEq, Eq, PartialOrd, Ord, Hash)]
 pub struct DownlevelLimits {}
 
+#[allow(unknown_lints)] // derivable_impls is nightly only currently
+#[allow(clippy::derivable_impls)]
 impl Default for DownlevelLimits {
     fn default() -> Self {
         DownlevelLimits {}
@@ -2776,7 +2773,7 @@ impl<L> TextureDescriptor<L> {
     /// # use wgpu_types as wgpu;
     /// let desc = wgpu::TextureDescriptor {
     ///   label: (),
-    ///   size: Extent3d { width: 100, height: 60, depth_or_array_layers: 2 },
+    ///   size: wgpu::Extent3d { width: 100, height: 60, depth_or_array_layers: 1 },
     ///   mip_level_count: 7,
     ///   sample_count: 1,
     ///   dimension: wgpu::TextureDimension::D3,
diff --git a/wgpu/src/util/init.rs b/wgpu/src/util/init.rs
index 4c5d6c3f70..32a0dd312a 100644
--- a/wgpu/src/util/init.rs
+++ b/wgpu/src/util/init.rs
@@ -2,23 +2,22 @@ use wgt::{Backends, PowerPreference, RequestAdapterOptions};
 
 use crate::{Adapter, Instance, Surface};
 
+#[cfg(any(not(target_arch = "wasm32"), feature = "wgc"))]
+pub use wgc::instance::parse_backends_from_comma_list;
+/// Always returns WEBGPU on wasm over webgpu.
+#[cfg(all(target_arch = "wasm32", not(feature = "wgc")))]
+pub fn parse_backends_from_comma_list(_string: &str) -> Backends {
+    Backends::BROWSER_WEBGPU
+}
+
 /// Get a set of backend bits from the environment variable WGPU_BACKEND.
 pub fn backend_bits_from_env() -> Option<Backends> {
-    Some(
-        match std::env::var("WGPU_BACKEND")
-            .as_deref()
-            .map(str::to_lowercase)
-            .as_deref()
-        {
-            Ok("vulkan") => Backends::VULKAN,
-            Ok("dx12") => Backends::DX12,
-            Ok("dx11") => Backends::DX11,
-            Ok("metal") => Backends::METAL,
-            Ok("gl") => Backends::GL,
-            Ok("webgpu") => Backends::BROWSER_WEBGPU,
-            _ => return None,
-        },
-    )
+    std::env::var("WGPU_BACKEND")
+        .as_deref()
+        .map(str::to_lowercase)
+        .ok()
+        .as_deref()
+        .map(parse_backends_from_comma_list)
 }
 
 /// Get a power preference from the environment variable WGPU_POWER_PREF
@@ -79,8 +78,7 @@ pub async fn initialize_adapter_from_env_or_default(
         None => {
             instance
                 .request_adapter(&RequestAdapterOptions {
-                    power_preference: power_preference_from_env()
-                        .unwrap_or_else(PowerPreference::default),
+                    power_preference: power_preference_from_env().unwrap_or_default(),
                     compatible_surface,
                 })
                 .await
diff --git a/wgpu/src/util/mod.rs b/wgpu/src/util/mod.rs
index f0a115a7c2..d50ee9796c 100644
--- a/wgpu/src/util/mod.rs
+++ b/wgpu/src/util/mod.rs
@@ -15,10 +15,7 @@ use std::{
 pub use belt::StagingBelt;
 pub use device::{BufferInitDescriptor, DeviceExt};
 pub use encoder::RenderEncoder;
-pub use init::{
-    backend_bits_from_env, initialize_adapter_from_env, initialize_adapter_from_env_or_default,
-    power_preference_from_env,
-};
+pub use init::*;
 
 /// Treat the given byte slice as a SPIR-V module.
 ///